Commonwealth Bank of Australia: Agentic AI System Detects Fraud Patterns, 80M Signals/Day, 20% Loss Reduction

Title

Commonwealth Bank of Australia: Agentic AI System Detects Fraud Patterns, 80M Signals/Day, 20% Loss Reduction

Content

Commonwealth Bank of Australia has deployed an agentic AI system that monitors more than 80 million signals per day across transactions, card payments, online payments, and digital banking interactions, contributing to a 20%+ drop in fraud losses in the first half of FY26 and automatically developing or updating 75% of the bank's card fraud rules. "The agent operates around the clock, continuously monitoring activity and adapting to emerging threats," said James Roberts, Executive General Manager of Fraud and Scams at Commonwealth Bank. The deployment is anchored in a fast-moving threat landscape: fraud and scam patterns in transaction and payments data evolve faster than static detection rules can keep up. "When suspicious patterns are identified, the system quickly assesses their severity, analyses context, and proposes new detection rules to help intercept them," Roberts explained. "The new agent goes beyond traditional AI by not only rapidly identifying new threats but also determining how it can seek to disrupt them." The investment is part of CommBank's $1 billion annual commitment to safeguard customers from fraud, scams, cyber threats, and financial crime. CommBank's in-house data science and engineering teams built the system in three months, with the underlying fraud platform already monitoring more than 80 million signals per day before the agent went live. The bank now processes more than 20 million payments on average each day and sends more than 40,000 proactive warning alerts to customers via the CommBank app, providing the real-time data substrate the agent operates on. Technically, the system is built on Snowflake's data cloud and powered by CommBank's cloud-based core banking platform, giving the agent seamless access to rich real-time data across payment and digital channels. Detection rules proposed by the agent are reviewed and approved by CommBank's fraud analytics team before deployment — a human-in-the-loop oversight process — and the bank notes that AI models across the business are governed by its risk management frameworks with clear human accountability for outcomes. At operational scale, the agent has already helped develop or update three quarters of CommBank's card fraud rules — the controls used to flag potential fraud — and the bank's fraud detection technology contributed to reducing fraud losses by over 20% in H1 FY26 compared with H1 FY25. "The technology allows us to identify unusual events in highly complex patterns of activity at far greater speed and scale, helping us detect emerging threats sooner and update our controls faster," Roberts said. Looking ahead, the deployment is positioned as part of CommBank's broader strategy to embed advanced AI into core operations, a stance the bank says is reflected in its ranking among the top four banks globally for AI maturity in the 2025 Evident AI Index.